Multitask Learning-Based Security Event Forecast Methods for Wireless Sensor Networks

Algorithm 1

) Input: data for all sub-regions   ()

() Parameters: and , main task and threshold

() Output: main task prediction data , sim tasks , and class loss

() Initialize:   ()  , arbitrary dimensional matrix ,

() iterate

()   for to m do

()      If

()        similarity = ,

()       If similarity ≥

()          = similarity,

()        With fixed and approximately solve for :

()       

()       Let

()      end if

()     end if

()  end for

()  Compute the singular value decomposition of :

()     (with diagonals of in descending order)

()  Let the rows of be the first rows of

() until converge

() using through CV output and